Current-Generating Double-Layer Shoe with a ... - Python Home Page

Mar 31, 2017 - transport limitations can dramatically reduce the generated current and power ... current.3 The task here is to produce the maximum pow...
0 downloads 11 Views 1004KB Size
Article pubs.acs.org/JPCC

Current-Generating Double-Layer Shoe with a Porous Sole: Ion Transport Matters Alexei A. Kornyshev,*,† Rebecca M. Twidale,† and Anatoly B. Kolomeisky‡ †

Department of Chemistry, Faculty of Natural Sciences, Imperial College London (University of Science, Technology and Medicine), London, SW7 2AZ London, United Kingdom ‡ Department of Chemistry and Department of Chemical & Biomolecular Engineering, Center for Theoretical Biological Physics, Rice University, Houston, Texas 77005-1892, United States ABSTRACT: Generating electrical current from mechanically forced variation of the contact area of electrode/electrolyte interface underpins one of the scenarios of harvesting electrical current from walking. We develop here theory of an electrical shoe with a porous sole with an account of both convection of the liquid electrolyte under pressure and ion migration with transmission-linetype charging of electrical double layer at the pore walls. We show here that ion transport limitations can dramatically reduce the generated current and power density. The developed theory describes the time dependence of the generated current and reveals its dependence on the main operation parameters, the amplitudes of oscillating pressure and frequency, in relation to the system parameters.



INTRODUCTION There is currently great interest in electrical current generation from mechanical motion of human individuals based on the principle of reverse electroactuation. Where direct electroactuation converts changing electric potential into mechanical motion,1,2 reverse electroactuation does the opposite, usually in the form of generating transient currents, with repeated oscillating motions producing ac current.3 The task here is to produce the maximum power density from the gadget at minimal degradation, i.e., maximal longevity of the device, and at low cost. There may be different principles of reverse electroactuation that we will not review here. The one we will focus on is based on mechanical changes of the electrical capacitance of the system. When the electrical capacitance is kept under constant voltage while connected to a battery, a change in this capacitance will cause a transient current in the network, charging or discharging the capacitor. A principle such as this can also be used in tactile sensing.4 The amount of current flowing in the system will be proportional to the change in capacitance, which in turn scales with its maximal value. Thus, manipulating the capacitance of a dielectric capacitor will produce much smaller currents than those using the electrochemical double-layer capacitors. Dielectric capacitance is inversely proportional to the thickness between the plates, the smallest value of which is usually not less than 100 nm, whereas electrochemical double layer capacitance is inversely proportional to the Gouy length, which depending on the concentration of electrolyte and voltage typically lies in the range of 1−10 nm. © 2017 American Chemical Society

The principle of capacitive reverse electroactuation using electrolytes has been patented by Krupenkin5 and developed in the pioneering work of his group.6,7 His devices are based on the compression of nonwetting electrolytic droplet(s) confined between two electrodes: When compressed, the two electrodes move closer together, and the droplet spreads and becomes wider, increasing its contact area with the electrodes, thereby increasing the capacitance. When the compression is removed, the droplets shrink back, and the contact area decreases reducing the capacitance. Since changing the capacitance at a constant voltage causes transient current, the alternation of compression and decompression will generate the alternating current patterns, and this is what Krupenkin’s group demonstrated experimentally. In a previous work,8 we developed a slightly different scenario of realization of the same idea. Instead of considering a flat capacitor system, we considered a porous electrode, ordinarily nonwetted with the electrolytic solution either due to its natural solvo-/iono-phobicity or due to the Cassie−Baxter condition9 of restraining of liquid penetration into a porous medium. Under sufficient external pressure, e.g., that emerging under a stepping foot, the liquid will penetrate the pores and the contact area between the electrolytic solution and the electrode will increase, thereby increasing the double layer capacitance. If the electrode is under a constant positive bias, then the electrical double layer will be rich in anions; the spreading of such a layer over the interior will produce a flow of Received: November 11, 2016 Revised: March 1, 2017 Published: March 31, 2017 7584

DOI: 10.1021/acs.jpcc.6b11385 J. Phys. Chem. C 2017, 121, 7584−7595

Article

The Journal of Physical Chemistry C

where r is the radius of the pore, k is a dimensional slip coefficient (which would have been zero for complete wetting conditions but is close to the pore radius for the nonwetting case), and η is the liquid viscosity.19 The expression for the evolution of charge in the pore due to the moving liquid and migration of counterions and co-ions in the flow of the liquid is obtained by incorporation of eq 1 into the transmission line equation. The transmission line theory describes the propagation of electrical current in the pore. Ideally, it works when the pore size is much larger than the thickness of the equilibrium electrical double layers at the pore walls so that one can speak about the ion transport in the “bulk” of the pore charging the double layer capacitor at the walls. Eventually, it will work even for narrow pores, but only if the electrical double layers on opposite sides of the pore do not overlap. In the transmission line theory, the ionic current along the pore axis is not conserved as it gets converted into the capacitor’s charge along the surface; in this respect, the electrode surface works as a “sink” of capacitive current. The transmission line model has been explored and exploited from the beginning of the 20th century (for review, see one of the latest papers utilizing it).14 Similar types of equations have emerged and been studied, not only in electrical engineering and electrochemistry but also in one-dimensional heat transfer.20 In the capacitive charging of a capillary of fixed length L completely filled with an electrolyte taking place after a voltage jump u̅, the equations of the transmission line theory give the following formula for the evolution of the net charge in the pore (for derivation, see Appendix A).

electrons away from the electrode to establish new charge equilibrium. As the electrolytic liquid recedes from the pore, the capacitance will diminish and cause a reverse flow of electrons. Under a set of simplifying assumptions, a fully analytical theory was developed to describe the operation of such a system. Among the assumptions made, the most important ones were as follows: (1) The flow of the liquid under the varying pressure obeys the Washburn equation. (2) There is no capillary hysteresis. (3) When the liquid flows into the pore, the double layer momentarily forms at the contact surface (adiabatic approximation). A full, detailed description of the rheology of the flow into narrow capillaries is a serious task, which must be further explored. However, for pores of 1 μm size, the Washburn equation can be a good semiphenomenological approximation.10−13 If the pore walls are ideally smooth, then hysteresis may not be essential. The role of ion transport limitations is the first thing to understand when going beyond the simple theory of the previous work.8 Indeed, will the double layer charging take place momentarily with the liquid flow? If not, what will be the effect on the generated current and the average power? The present work is intended to answer these questions by investigating a model of ion transport into the liquid that moves inside the pore. We have chosen the simplest possible model that integrates the transmission line theory of ion transport into the Washburn-like description of a pressure-induced liquid flow. This simplification allows us to develop a fully analytical theory, which will recover the result of the previous work as particular limiting case for very fast migration of ions. The resulting formulas for the time-dependent current and the average power will allow us to understand what is needed to minimize the ion migration limitations and the characteristic time related to them. Ion transport effects in “sister problems”, such capacitive ionization or more generally capacitive salinity gradient energy systems, have been intensively studied in a number of works, see e.g., Biesheuvel et al.,14 some papers in this journal,14−17 and a review article,18 but to our knowledge, they have not been explored in porous reverse electroactuators.

Q (t ) = uC ̅ 0lpL

(1 − sin(ωt ) , t ≥ π /2ω

τ=

2 2 ⎧ −(n + 1 ) π 2 l 2 · τt ⎫ 2 ⎨ L ⎬ 1 e − 2 ⎭ ⎩

(n + 12 )

C0l Σ

(4)

where ∑ is the ionic conductivity of the electrolytic solution that we use to fill the pore. This time increases with poorer conductivity of the electrolyte because the ions will be migrating slower to reach the areas they need to charge and as the double layer capacitance becomes larger because for a given voltage more charge will need to be transported to those areas. However, in a transmission line system the relaxation time is much longer than τ; it is, actually, equal to

(1)

Here P0 is the amplitude and ω is the frequency of the oscillating external pressure, e.g., emerging from walking, and t is the time over which the charging takes place. The parameter A (m2 Pa−1 s−1) is defined as

r 2 + 4kr A= 8η

n=0

1

Here, u̅ is the potential difference between the electrode (pore surface) and the bulk electrolyte, C0 is the double layer capacitance per unit surface area of the pore (considered to be voltage independent), lp is the average pore perimeter, and 1/l is the specific surface area of the pore, i.e., 1/l = S/V where S is the surface area of the pore and V is the volume of the pore. For cylindrical pores, we have lp = 2πr and l = r/2; therefore, lp = 4πl. This relationship is used throughout this work when calculating the values of parameters used in the graphs. The following combination of parameters determines here the quantity of the dimensionality of time.

MODEL AND BASIC EQUATIONS For this model, we will investigate charging of a pore using the transmission line theory in order to account for the redistribution of ions in an electrolyte not being instantaneous when following the moving liquid. We will develop the simplest version of the theory, taking into consideration that the penetration depth of the liquid into the pore, L(t), changes with time. For this quantity we will use the form derived in the previous work,8 obtained for a periodical change of external pressure and the use of Washburn equation, which describes how the liquid moves inside the pore under a variable pressure: 2AP0 ω





(3)



L (t ) =

2 π2

τp =

1 L2 τ π 2 l2

(5)

This time increases with the length of the filled pore, as L2, and decreases with increasing pore cross section.

(2) 7585

DOI: 10.1021/acs.jpcc.6b11385 J. Phys. Chem. C 2017, 121, 7584−7595

Article

The Journal of Physical Chemistry C Note that eq 3 is valid under the assumption that C0 does not depend on applied voltage. This is generally never the case in electrochemistry, but this approximation allows us to obtain analytical results; furthermore, the effect of voltage dependence often averages out in the whole sample when considering pore size distribution.21 In our case, however, we are not considering a voltage jump. On the contrary, the electrode potential relative to the bulk electrolyte (reference electrode) is kept constant, and it is now the depth of the liquid penetration into the pore which changes. However, all the boundary conditions are the same. If we assume that the speed of migration of ions is faster than the convection of the liquid, then we can suggest the next level of the “adiabatic” approximation by substitution of eq 1 for L(t) into the transmission line equation (eq 3). By using this crude approximation, we decouple the problems of convection and migration but incorporate them in one process. We critically analyze it in the end of this section. Once adopting it, we obtain a formula for the evolution of the net charge in the pore over one period of oscillation of external pressure

extrapolation, which as we will see below seems to give physically reasonable results. With these reservations in mind, let us now explore the consequences of eq 6.



RESULTS AND DISCUSSION Charge Evolution. In Figure 1, using a set of estimated realistic values for the parameters listed in the caption, we show

2AP0 (1 − sin(ωt )) ω 2 ⎡ 2AP ⎤⎫ ⎧ −(n + 1 ) π 2l 2 / ⎢ ω 0 (1 − sin(ωt )) τt ⎥ 1 2 ⎣ ⎦⎬ , π ⎨ − 1 e 1 2⎩ ⎭ n+ 2

Q (t ) = uC ̅ 0lp ∞

2 ∑ π 2 n=0

(

)









/2ω ≤ t ≤ 5π /2ω

Figure 1. Charge accumulated in a solvophobic pore over a period of application of periodically varying external pressure: effect of ion transport limitations. Different curves correspond to different values of the intrinsic migration times τ. The slower the characteristic migration time τ (see text), the less charge will have time to accumulate in the pore. The values for the parameters used in this plot are u̅ = 0.3 V, C0 = 0.1 F m−2, l = 0.5 · 10−6 m, A = 10−10 m2 Pa−1 s−1, P0 = 105 Pa, and ω = 2π s−1. The effect of the characteristic migration time on charge accumulation in the pore is investigated here over three different orders of magnitude.

(6)

This equation describes the charge within each period. At the onset point of each period there is no liquid in the pore, L = 0, and this is the same at the end of the period. One of the main assumptions here is that when the liquid returns from the pore to the bulk there is no memory in the system: counterions do not crowd up at the entrance of the pore. Thus, hereafter, we will be considering only one period of pressure oscillation, assuming that once the electrolyte has been pushed out of the pore to its original position there will be no “memory” of the previous period. The electrolyte will act in exactly the same way every single time it enters the pore. Looking at eq 6, we see that the characteristic time to charge the pore over the maximum penetration length for a given amplitude of pressure is not τ, as we have already commented above, but something close to τp =

8AP0 π 2l 2ω

τ

the evolution of charge within one pore over a 1 s period of pressure oscillation for indicated values of the intrinsic migration times. There are three regimes in the dynamics of the charge accumulation as shown in Figure 1. The amount of charge initially increases linearly, the accumulation rate then changes to produce a shallower gradient, and finally it starts to linearly decrease. From eq 6, it can be shown that for times close to

(

π

)

π/2ω we have in the first regime Q (t ) ≈ C0lp AP0ω t − 2ω , while in the third regime closer to 5π/2ω,

(7)

How large could this time be? Taking Σ = 0.14 S/m for 0.1 molar aqueous solution of KCl, l = 5 × 10−7 m, and C0 = 0.1 F/ m2, we get τ = 3.6 × 10−7 s. With A = 10−10 m2 Pa−1 s−1 and P0 = 105 Pa, this gives τp = 1.9 s, which is longer than the 1 s period of oscillation. For 1 M electrolyte, the conductivity will be 10 times larger, and τ and τp will be 10 times shorter. In that case, our adiabatic approximation would be rigorous, but then the effect of ion migration limitations will be small. However, there is in fact no simple relaxation time behavior in eq 6, because of the factor 1 − sin(ωt) in the denominator of the exponent. At certain sections of a period when sin(ωt) is close to 1, the response is much faster. Furthermore, due to the series structure of the result, the time dependence at small times is not exponential but rather proportional to √t (c.f. Appendix A). Thus, extending the theory to the case when τp is on the order of the period of pressure oscillation has a character of

Q (t ) ≈ C0lp AP0ω

( 25ωπ − t).

The charge evolution in the middle regime is different and is already affected by the migration time. This behavior is nontrivial, but is still easy to rationalize. At the beginning and the end of the period, the chargingdischarging dynamics is independent of the ion migration because in these cases the rate-limiting is the motion of the liquid as a whole. It is only in the middle region, when the ion migration cannot follow the liquid filling or leaving the pore, that the diffusion of ions becomes a rate-limiting step. Therefore, the faster the ion migration, the shorter the duration of this intermediate regime. In other words, as the liquid enters the pore, the ions immediately follow it because the speed of the liquid motion in the pore is initially low, as well as the penetration depth being small. Later, however, the speed of

7586

DOI: 10.1021/acs.jpcc.6b11385 J. Phys. Chem. C 2017, 121, 7584−7595

Article

The Journal of Physical Chemistry C pore filling accelerates, and the depth increases. At the central point of the period (t = 3π ), the penetration depth reaches its 2ω maximum and begins to return to the opening. The peak of the graph is the point at which the flow of counterions meets the returning liquid front, and the liquid and the ions together return to the bulk liquid outside the pore. The migration time, τ, is short and, depending on the conductivity of the electrolyte and the double layer capacitance per unit surface area, lies in the range of 10−7−10−5 s. The characteristic time to charge the pore, τp, over the maximum penetration length, however, is much longer. For a set of typical parameter values, l = 0.5 × 10−6 m, A = 10−10 m2 Pa−1 s−1, P0 = 105 Pa, ω = 2π s−1, and τ = 10−6 s, eq 7 gives us τp = 1.29 s. Such a value is close to the pressure oscillation period (1 s); hence, we should see a delay in “charge delivery” into the pore sensitive to the values of parameters − the exact value of τ in the first place. By changing the order of magnitude of τ, we can see what will be the effect of the delayed migration of ions. When comparing the characteristic time, τp, for the three different values of τ used in the plots above, we obtain τp = 0.129, 1.29, and 12.9 s when τ = 10−7, 10−6, and 10−5 s, respectively. From these values and applying them to eq 6, we can deduce that as the characteristic time, τ, becomes quicker, the exponential term vanishes much faster, bringing the result much closer to that of the adiabatic approximation with no lag in the migration current. This makes sense, as the quicker the ions migrate and form the double layer inside the pore, the faster the current will be generated and the smaller the lag in the current. Generated Current. The net current of ions is just time derivative of the charge J (t ) =

d Q (t ) dt

Figure 2. Absolute current generated in a single solvophobic pore over a period of application of periodically varying external pressure. As in Figure 1, the slower the intrinsic migration time τ (see the text), the less time the charge has to accumulate in the pore, so less current will be generated per period. Therefore, for a slower characteristic time, the plot produced a more prominent lag in the current. The values for the parameters used in this plot are the same as those used in Figure 1. The effect of the characteristic migration time on current generation in the pore is investigated here over four different orders of magnitude, recovering the result of the previous work8 only when the migration time reached is unphysically small (10−8 s).

very short migration time (unphysically short), to check whether we could recover the fully adiabatic regime limit.8 Similar to the charge dynamics, there are three regimes for the current. At the beginning of the period, the current is constant and is equal to J = C0lp AP0ω , while at the end of the period it has the same magnitude but opposite sign. As we see from Figure 2, changing the intrinsic migration time τ will change the way the charge evolves and therefore change the amount of current the electrode will produce over one period. As τ increases, the current is generated only at the beginning and the end of the period, remaining very small in the middle. As τ decreases, current generation is more even; in this process, ions with almost no delay charge the length of the pore penetrated by the liquid. When the order of magnitude of τ becomes as small as 10−8, the plot does not show any delay, essentially reproducing the result obtained in the previous work under the adiabatic regime.8 Power. Calculation of the power that an electromechanical capacitive current generator can produce is a special task with requires consideration of particular circuit which may generally contain resistive, inductive, and capacitive elements. The combination of these elements will affect the value and time dependence of the generated current, as well as the power pumped into the load. We consider nothing of the kind in this paper, but in Appendix C, we investigate the conditions under which we can calculate the capacitively generated current neglecting the load. Thus, when speaking below about the “power” we will estimate the maximum amount of electrical energy that our variable capacitor can produce over half-aperiod of pressure oscillation in a “short-cut” circuit. Hereafter, we will designate it as “short-circuit” power density, and when the adjective is omitted for brevity, we still mean such quantity. It benchmarks the general capability of device, the implementation of which must generally be studied with the

(8)

It will be equal and opposite in sign to the electronic current in the electrode to meet the ions, but since the current in electrical engineering is defined as the flow of positive charges, eq 2 gives us the current in the network. Note again that the ion current is not constant along the pore; it “sinks”, charging the pore walls. By definition the ionic current must be zero at the front of the liquid moving in the pore, as no ion can cross this front. Differentiation of eq 4 gives the time dependence of the net ionic current inside a single pore J (t ) =

2ω ⎧ −sin(ωt ) + ωt cos(ωt ) + 1 ⎫ ⎬ ⎨ AP0 ⎩ (1 − sin(ωt ))2 ⎭

2 uC ̅ 0lpl τ



1 − sin(ωt )

1

2

⎡ 2AP0 t⎤ ω (1 − sin(ωt )) τ ⎥⎦

∑ e−(n+ 2 ) π l /⎢⎣ 22

n=0



2uC ̅ 0lp π2

ωAP0 2



cos(ωt ) (1 − sin(ωt ))

∑ n=0

1 2

(n + 12 )

2 ⎡ 2AP ⎤⎫ ⎧ − n + 1 π 2l 2 / ⎢ ω 0 (1 − sin(ωt )) τt ⎥ ⎣ ⎦⎬ , ⎨1 − e ( 2 ) ⎩ ⎭ ⎪

π /2ω ≤ t ≤ 5π /2ω



(9)

with all parameters having the same definition as those in eq 6. Figure 2 plots this current for three values of the intrinsic migration time, τ, the same as those used in the plots for the charge evolution in Figure 1, plus one more curve, plotted for a 7587

DOI: 10.1021/acs.jpcc.6b11385 J. Phys. Chem. C 2017, 121, 7584−7595

Article

The Journal of Physical Chemistry C

For a typical set of parameters, l = 0.5 × 10−6 m, A = 10−10 m Pa−1 s−1, P0 = 105 Pa, τ = 10−6 s, and

consideration of the full circuit (for details, see Appendix C). For such a defined quantity, we can obtain a simple analytical expression. This can be calculated as

W = ⟨J ⟩u ̅

2

x=

ε ω πr 2 π

3π /2ω

∫π /2ω

J (t ) d t

= 9.375 × 10−3, eq 14 gives f = 0.806 so that due

to the lag in the “ion response” more than 80% of the power would be lost compared to the fully adiabatic “ideal world”.8,22 However, if the migration time is 1 order of magnitude shorter, then the same estimate will give f = 0.392, i.e., only 39.2% of power is lost due to ion migration limitations. On the contrary, if migration time is estimated to be 10 times longer, f = 0.939, the system will function at only 6% of its ideal efficiency. We show the graph of the power (Figure 4), for the same set of parameters, showing its dependence on τ.

(10)

where the average current density J is defined as characteristic current per unit surface are of the sole of the shoe, averaged over a half of a period (because two halves of the period are symmetric). ⟨J ⟩ =

3l 2 8AP0τ

(11)

Here, ε is the porosity of the sole of the shoe and r is the radius of a single pore (we mean the average radius). Recalling eq 8, we can remove the integral in eq 11 to obtain ⟨J ⟩ =

⎛ π ⎞⎤ ε ω ⎡ ⎛⎜ 3π ⎞⎟ ⎢⎣Q ⎝ ⎠ − Q ⎝⎜ ⎠⎟⎥⎦ 2 2ω 2ω πr π

(12)

Substituting eq 6 into eq 12 and taking into account that 2 π2



(

∑n = 0 n +

1 −2 2

)

= 1, we finally obtain a compact ex-

pression for the power density per unit area of the sole W=ε

⎛ ⎛ 3l 2 ⎞⎞ 2 C0u ̅ 2 AP0ω ⎜⎜1 − f ⎜ ⎟⎟⎟ πl ⎝ 8AP0τ ⎠⎠ ⎝

(13)

where the function f(x) is defined as f (x ) =

2 π2



∑ n=0

1

1

1 2 2

(n + )

2

e−(n + 2 ) π x 3

(14)

Figure 4. Generated power for the charging of hydrophobic pores, changing the intrinsic migration time of ions in the electrolyte, plotted using eq 13. At very short ion migration times, the generated power density increases slowly with decreasing migration time. However, as the migration time becomes longer, the power density decreases much faster, tending toward a much lower power density. The parameters used in this plot are the same as those in Figure 1, with ε = 0.5. The characteristic migration time, τ, is changed logarithmically to view the full effect on the power density.

Without the second term in the brackets, eq 13 would give the same result as in the fully adiabatic theory of the previous work8 (see eq 17 of ref 8, with r there replaced by 2l, with account to the correction to this equation).22 Thus, it is the f term that reduces the power density. Of course, when τ → 0, then x → ∞ and f → 0, and the result of the fully adiabatic theory8,22 is recovered. Otherwise, the f term may comprise a substantial reduction of the power density. Since f(x) is a universal function and not dependent on any other parameters, it is worth displaying (Figure 3).

In accordance with the above-discussed analysis of the f factor, Figure 4 shows that there is a point around τ = 10−7 where power density increases at a slower rate with decrease of the rate of migration. This is because for the taken set of parameters the ions are reaching the value of τ where they have time to reach even the peripheral parts of the liquid in the pore while the liquid moves there; the plot looks as though it will reach a saturation point if the migration time is decreased even further. We can see, however, that the general trend is that as the migration time increases the average current output over the soles decreases and so does the power density. Changing the Period of Oscillation at Constant Migration Time. So far we have kept all parameters constant aside from the characteristic migration time, τ. Now we will investigate the effects of changing the period of walking, ω, while keeping the migration time at a constant value of 10−6 s. Realistically, a rate of 1 step per second is a slow walk, so we will consider what happens to the generated power density when we increase the speed, or decrease the length of the period, to 2 steps per second; ω = 4π s−1 (a fast walk); and 3 steps per second; ω = 6π s−1 (a fast run). In in Figures 5 and 6,

Figure 3. Graph of f function defined by eq 12. 7588

DOI: 10.1021/acs.jpcc.6b11385 J. Phys. Chem. C 2017, 121, 7584−7595

Article

The Journal of Physical Chemistry C

overall amount of charge generated decreases. This is due to the speed of the oscillating pressure increasing with decreasing period so that the liquid enters the pore much faster, but it doesn’t have enough time to go deep into the pore. To recharge the double layers, the ions have to migrate shorter distances, and the lag effect is less felt. This is supported by the trends in the current shown in Figure 6. As the period gets shorter, the initial current generated at the pore becomes larger. The combined effect of the frequency and period on the power density generated in the pore is simple. Following eq 13, it scales ∝√ω, in the same way as it does in the fully adiabatic case,8,22 but due to the migration delay, the proportionality factor is smaller. However, again, we can see that the faster you walk the greater the amount of the generated power. This trend is shown in Figure 7.

respectively, we show the effect that the frequency has on the evolution of charge in the pore and the generated current

Figure 5. Charge accumulated in a solvophobic pore over one period of application of a periodically varying external pressure. The plot shows the effect of frequency of the applied pressure on the charge accumulated in the pore. As the frequency increases from ω = 2π s−1 to ω = 6π s−1, the total amount of charge generated at the pore decreases. This is because as the frequency increases the liquid spends less time in the pore, so the ions have less time to form the double layer inside the pore. The parameters have the same values of those used to plot Figure 1, aside from τ being fixed at 10−6 s and ω being set as shown in the inset.

Figure 7. Short-circuit power density generated with changing frequency of oscillating pressure, shown for three different migration times. The power generated increases with the square root of frequency, described by the simple power law of eq 13 As migration time decreases, the power increases. The parameters used in this plot are the same as those in Figure 1, aside from ω now continuously varying between 2π and 6π s−1 and ε = 0.5.

Decreasing the size of the pores allows a larger amount of current to be generated due to the increase in the number of pores on the sole per unit cross-sectional area; therefore, the system will operate with larger overall surface area available for charging. Charging of all pores will proceed in “parallel” so that that the increase of the area in this way will not slow down charging. Pressure Dependence. Here the f term brings a dramatic effect. The behavior of the power with the increase of pressure is determined by the behavior of the function 1 − f(x). Indeed, there is an asymptotic law for this function: As x → 0, 1 − f(x) 1 ≈ 2√x, but the factor in front of 1 − f(x) in eq 13 is ∝ x .

Figure 6. Absolute current generated in a single solvophobic pore over one period of application of periodically varying external pressure. This plot shows the effect of the frequency of the applied pressure on the generated current. The parameters used are the same as those in Figure 1, aside from τ being fixed at 10−6 s and different values of ω being set between 2π s−1 and 6π s−1, as shown in the inset.

1 − f (x)

Thus, as x → 0, x ≈ 2 . With increasing P0, x → 0, and the power increases. However, it then saturates to a constant value. This value is given by a simple formula:

In Figure 5, we can see that changing the frequency does not change the shape of the plot qualitatively, but it affects it quantitatively. Of course, the period of a single step, T, 2π decreases as frequency, ω, increases: T = ω . Thus, as the period decreases, the plots become narrower. The initial gradients increase, which indicates that speed of charge generation increases with decreasing time period, even if the

W = 0.78εC0 U̅ 2

ω τ

(high pressure limit)

(15)

Of course when pressure goes to zero, f(x) ≈ 0, and 1 − f(x) → 1. Thus 7589

DOI: 10.1021/acs.jpcc.6b11385 J. Phys. Chem. C 2017, 121, 7584−7595

Article

The Journal of Physical Chemistry C W=ε

2 C0 U̅ 2 AP0ω πl

(low pressure behaviour)

and independent of convection but proceeding in a liquid column of varying length. We considered the principle issues of this process and thus studied effects in one single pore, extending it later on a monodisperse system of parallel pores. Integration of our findings into the theory of electrodes with pore size distribution and complex shapes of the pores was beyond our goals; such investigation may have multifarious aspects and will be the subject of future studies. However, modern technology can build at least laboratory prototypes in which the electrode may have a system of monodisperse cylindrical pores with an ideal internal surface so that the results of our findings can be tested experimentally as they are. Using the chosen theoretical model, we have found that the ion migration limitations may have a serious impact on generated power. It affects the slope of the frequency dependence of generated power, which as a function of frequency should obey the ω1/2 law. It levels off its pressure dependence; with the increase of the amplitude of the oscillating pressure, the power first increases square root wise, ∝P1/2 0 , but then reaches saturation. The latter takes place only when the ion migration limitations are taken into account. We should recall that the simple combination of two theories seems physically justified when the characteristic time τp (eq 7) is not much greater than the oscillation period T = 2π/ω. Eventually with the increase of frequency, τp, which is ∝1/ω, decreases with the same rate; thus, the ω1/2 law is, perhaps, not limited by the approximations of the theory. We are less sure about the pressure dependence law. If the

(16)

(which is the result of the previous work8 with account for correction of the numerical coefficient).22 Thus, the power increases initially as the square root of pressure amplitude and then saturates at the maximally possible limiting value. This limitation holds because increasing pressure amplitude decreases the time that liquid spends in the pore, and there is no time for redistribution of ions to take place. Figure 8

criterion is indeed

Figure 8. Pressure dependence of the short circuit power density, as a function of p, the portion of pressure of 105 Pa. (For an average adult individual, p ∼ 1). The blue solid curve is calculated via eq 13. Parameters are the same as those in Figure 1, and ε = 0.5. For comparison, the dotted green curve corresponds to the case of no diffusion limitations (eq 16).

2AP0 22

π lω

τ